hi there 👋🏽, i'm

Wahid Mahmood

A Machine Learning Engineer specializing in both cutting-edge research and intelligent system development that solve real world problems.

Hire Me Download CV

About Me

I am a dedicated and dynamic Machine Learning Engineer at ReliSource, driven by curiosity and a commitment to continuous growth.

Passionate about learning, I consistently strive to enhance my skills and stay at the forefront of innovation. I thrive on challenges and approach problems with enthusiasm and determination.

Beyond my professional expertise, I value effective communication and collaboration. I also enjoy intellectual pursuits such as playing chess and watching football, which help me relax and grow personally.

Educations

Khulna University of Engineering & Technology (KUET)

B.Sc in Computer Science and Engineering (CSE)

---------------------------------------------------------------------------------------

CGPA: 3.53/4.0

Natural Language Processing
Statistics
Data Mining
Image Processing and Computer Vision
Artificial Intelligence
Linear Algebra
Database System
Operating System

Research Work

Thesis and Publications

Thesis

Study on Agricultural Land Reduction in Bangladesh using Satellite Images

Wahid Mahmood
Dr. Sk. Md. Masudul Ahsan

Collected and preprocessed raw remote sensing image over the area of Bangladesh from Sentinel-2 satellite

Created agricultural land change detection dataset consisting over 500 image pairs over the area of Bangladesh.

Explored multiple ML techniques (XGBoost, SVM) and extended the open sourced framework called Open-CD

Github →

Paper 1

A Semi-Supervised Approach For Detecting Agricultural Land Reduction In Bangladesh Using Satellite Images

IEEE International Conference on Computing, Communication, and Intelligent Systems (COMPAS), 2024. DOI: 10.1109/COMPAS60761.2024.10795941

Wahid Mahmood
Dr. Sk. Md. Masudul Ahsan

Developed a novel semi-supervised approach for agricultural land change detection, which significantly reduced the time and hardship required for manual annotation.

Further evaluated the created dataset with other deep learning methodologies.

Presented poster on this topic.

Paper 2

NSNUNet: A Neighbourhood Attention Based Siamese Nested Unet for Change Detection using Satellite Images

IEEE International Conference on Computer and Information Technology (ICCIT), 2024.DOI: 10.1109/ICCIT64611.2024.11022454

Wahid Mahmood
Taslima Akhter
Dr. Sk. Md. Masudul Ahsan

Proposed a novel neighborhood attention based UNet architecture, improving accuracy and efficiency in remote sensing change detection.

The proposed model was evaluated on both low resolution and high resolution image dataset, outperforming other methodologies on both

Work Experience

ReliSource

Machine Learning Engineer

June 2024 - Current

AI Chatbot: RAG based customer service AI Chatbot using Langchain, FastAPI and OpenAI, which achieved 4-second response time, including source citation and message streaming by implementing efficient retrieval and answer generation workflows.

Medical Inventory Demand Forecasting: An effective medical inventory demand forecasting pipeline automates the prediction of medication needs using limited historical time-series data. Enhanced operational efficiency using multi step approach including temporal analysis, feature engineering and finally utilize agentic AI to generate actionable restocking recommendations.

Drug Discovery Pipeline: ML-based drug discovery pipeline to screen and validate potential drug candidates including ADMET analysis, utilizing tools like AlphaFold, DeepChem, Biopython, RDKit etc.

PDF Parsing: Created a PDF parsing pipeline, which will extract necessary information from pdf using Data Embedding and LLM and store it on excel

Projects

Some of my works

01

Intent Based Product Search

A product search system for e-commerce sites which can detect the user’s intent using search history and query optimization and provide the most optimal search results.

Used finetuned model of llama 3.2 1B instruct and hybrid search for retrieving products with 2 sec latency.

Langchain
FastAPI
PGVector
Ollama
Redis
Langsmith

Github →

02

Dokan Koi?

A Flutter-based e-commerce app connecting users with the best nearby stores and stores with all the customers of the area.

Features include shop location tracking, real-time updates, and notifications by utilizing Firebase for data storage, authentication, and file storage.